On Wed, 31 Mar 1999 15:09:02 -0500 
 Bill Sommerfeld <sommerfeld@orchard.arlington.ma.us> wrote:

 > If we have effective hard limit support in the pool code, is there
 > still a good reason to put mbufs in a separate map on some platforms?

Yes, because that submap (mb_map for clusters, kmem_map for mbufs) is
protected w/ splimp(), and kernel_map is not.

 > What makes the alpha and mips special?

They use a direct-mapped segment (KSEG) to `map' the single page pool
pages, and thus don't need to use the TLB for pool items.

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>